Humanoid robot joint torque sensor chip control method and system

Abstract

The application discloses a humanoid robot joint torque sensing chip control method and system, comprising: collecting joint torque, rotation angle displacement and acceleration data and transmitting to a control platform, constructing a flexible joint torque recursive model to deduce torque transmission characteristics, using a harmonic resonance algorithm to suppress harmonic resonance interference in torque transmission, realizing distributed torque allocation through a cooperative allocation algorithm, combining feedback data to correct control parameters and driving the joint to execute actions to form a closed-loop control link; the system corresponds to the method to build an integrated control architecture. Through dynamic linkage and data synchronous transmission of each link, the method and system strengthen the adaptability of sensing data and control process, standardize the torque distribution rules of multiple joints and multiple robots, improve the accuracy, dynamic adaptability and cooperativity of joint control, effectively solve the problems of module dispersion and response lag in traditional technology, and are suitable for humanoid robot joint torque control in complex operation scenarios.

Description

Technical Field

[0001] This invention relates to the field of robot joint control technology, and more particularly to a method and system for controlling joint torque sensing chips in humanoid robots. Background Technology

[0002] Humanoid robots are rapidly evolving towards multi-scenario adaptability, high dynamic response, and collaborative swarm operation. Joints, as the core hub connecting the robot body and actuators, directly impact the robot's overall motion performance through the accuracy of torque control, dynamic adaptability, and multi-unit collaboration. In complex operating environments, the joint torque transmission process is susceptible to factors such as multi-component coupling, dynamic interference from drive components, and collaborative scheduling in swarm operations. Traditional control modes struggle to achieve efficient linkage between data acquisition, torque adjustment, interference suppression, and collaborative control, failing to meet the precise control requirements of robots in diverse motion states. As a core component for torque sensing and control command execution, joint torque sensing chips need to be efficiently adapted to multi-joint collaborative control systems and swarm management platforms to construct a full-link control architecture from data acquisition to command execution, addressing the challenges of dynamic torque adjustment in complex motion scenarios.

[0003] Existing technologies in humanoid robot joint torque control suffer from two major shortcomings: First, they lack a deeply integrated architecture encompassing sensor chips, multi-joint torque adjustment mechanisms, and cluster control platforms. The dispersed operation of each control module leads to timing discrepancies in data transmission and command execution, making it difficult to match torque adjustment parameters with joint dynamic states in real time and hindering multi-stage dynamic linkage control. Second, the torque-related data collected by sensor chips is poorly adapted to subsequent control processes. There is a lack of dedicated data processing and parameter calibration mechanisms for joint dynamic changes. Furthermore, in multi-robot collaborative scenarios, the lack of a unified torque distribution coordination mechanism results in lag in control response, insufficient torque allocation rationality, and an inability to fully leverage the real-time sensing capabilities of sensor chips and the adjustment potential of the control system, thus restricting the overall performance improvement of humanoid robot joint control. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of the existing technology, the present invention provides a method and system for controlling the joint torque sensing chip of a humanoid robot.

[0005] The technical solution adopted in this invention is a control method for a humanoid robot joint torque sensing chip, comprising the following steps: S1, collecting real-time torque sensing data, joint rotation displacement data, and joint motion acceleration data of each joint through a humanoid robot joint torque sensing chip, and transmitting the collected data to a robot cluster collaborative intelligent control platform for data parsing and correlation mapping; S2, constructing a flexible joint torque recursive model based on the parsed joint data, and using this model to perform hierarchical deduction of the dynamic changes in the joint torque transmission path to determine the transmission characteristics and coupling relationship of joint torque in a multi-link structure; S3, employing a series elastic actuator resonance algorithm to mitigate resonance interference during the joint torque transmission process. The signal is dynamically suppressed by adjusting the stiffness matching parameters of the actuator's elastic element through an algorithm to achieve a dynamic balance between torque transmission and resonance suppression; S4, the multi-joint torque collaborative allocation algorithm is used to distribute the derived joint torque data in a distributed manner, and the torque allocation weight is determined according to the motion state and load requirements of each joint; S5, the allocated torque control signal is transmitted to the humanoid robot joint actuator through the robot cluster collaborative intelligent control platform, and the torque control parameters are dynamically corrected by combining the real-time feedback data from the sensor chip; S6, the joint actuator is driven to complete the specified motion action based on the corrected control parameters, and the joint torque response data is continuously collected through the sensor chip to form a closed-loop control link.

[0006] Furthermore, the recursive model expression for the flexible joint torque is as follows: ,in, For the first The recursive torque value of each joint, Let be the initial torque value of the i-th joint. Let be the rotation angle of the i-th joint. Let be the stiffness coefficient of the series elastic actuator. For the deformation of the elastic element, Let be the rotational inertia of the i-th joint. Let be the angular acceleration of the i-th joint. This is the torque transmission coupling coefficient.

[0007] Furthermore, the expression for the series elastic driver resonance algorithm is as follows: ,in, The transfer function after resonance suppression. The resonant frequency, For complex frequency variables, The damping coefficient is... For damping feedback coefficient, For the moment of inertia of the motor, This is the motor damping coefficient. This is the frequency compensation factor.

[0008] Furthermore, the expression for the multi-joint torque collaborative allocation algorithm is as follows: ,in, The distributed joint torque, The total number of joints. Assign weighting coefficients to the torque of the i-th joint. This is the joint motion sensitivity coefficient. Let be the generalized coordinate of the i-th joint. For generalized coordinate changes, For load adaptability coefficient, Let be the load torque of the i-th joint. To coordinate the allocation of coordination factors.

[0009] Furthermore, the torque control parameter correction model expression of the robot swarm collaborative intelligent control platform is as follows: To control the parameter correction amount, This is the proportionality coefficient. The integral coefficient is... The differential coefficients are... For coupling correction coefficients, For reference torque value, This is the feedback torque value of the sensor chip. This is a multi-joint coupling torque.

[0010] Furthermore, the signal acquisition model expression of the humanoid robot joint torque sensing chip is as follows: ense ,in, The chip outputs a voltage signal. ense is the sensitivity coefficient of the sensor chip. This represents the actual torque of the joint. This is the torque-to-voltage conversion factor. The rate of change of torque, For dynamic response coefficients, Zero-point offset voltage, This is the environmental compensation coefficient.

[0011] Further, step S3 includes the following sub-steps: S31, extracting the resonant characteristic frequency from the torque data output by the flexible joint torque recursive model, and determining the amplitude and phase characteristics of the resonant signal through spectrum analysis; S32, based on the structural parameters and motion state of the series elastic actuator, adjusting the stiffness matching parameters and damping feedback coefficients in the algorithm to construct a resonant suppression parameter set; S33, inputting the resonant characteristic parameters and suppression parameter set into the series elastic actuator resonant algorithm, and generating a resonant suppression control signal through algorithm calculation; S34, superimposing the resonant suppression control signal onto the joint torque control link to cancel the resonant interference in the torque transmission process in real time, and simultaneously verifying the effect by collecting the suppressed torque data through a sensor chip.

[0012] Further, S4 includes the following sub-steps: S41, acquiring real-time load data, motion speed data, and joint health status data of each joint through the robot cluster collaborative intelligent control platform, and establishing a joint status evaluation matrix; S42, determining the torque bearing capacity and motion priority of each joint based on the joint status evaluation matrix, and assigning corresponding torque allocation weight coefficients and load adaptation coefficients; S43, inputting the weight coefficients, load data, and recursive torque values ​​into the multi-joint torque collaborative allocation algorithm, and dynamically allocating torque data through the distributed operation of the algorithm; S44, outputting the allocated torque control commands for each joint, and simultaneously recording the parameter adjustment data during the allocation process to provide a basis for subsequent control optimization.

[0013] Further, S5 includes the following sub-steps: S51, receiving the torque control command output by the multi-joint torque collaborative allocation algorithm, and determining the command execution priority by combining the global control strategy of the robot cluster collaborative intelligent control platform; S52, transmitting the torque control command to the drive module of each joint actuator through the communication link, and simultaneously activating the high-frequency data acquisition mode of the sensor chip; S53, comparing the real-time torque data fed back by the sensor chip with the command torque data, and calculating the torque control deviation value; S54, based on the deviation value and the joint motion state, dynamically correcting the torque control parameters through the control platform, and retransmitting the corrected command to the drive module for closed-loop control.

[0014] A humanoid robot joint torque sensing chip control system, applied to the control method of humanoid robot joint torque sensing chips, includes: a multi-dimensional joint torque sensing and acquisition unit for acquiring torque data, angular displacement data, and acceleration data of each joint, establishing a bidirectional data transmission link with the robot swarm collaborative intelligent control platform; a flexible joint torque hierarchical recursive calculation unit, connected to the humanoid robot joint torque multi-dimensional sensing and acquisition unit, receiving the acquired data and performing torque transmission characteristic deduction through a preset flexible joint torque recursive model; a series elastic actuator resonance dynamic suppression unit, connected to the flexible joint torque hierarchical recursive calculation unit, suppressing resonance interference in the torque data through a series elastic actuator resonance algorithm; and multi-joint torque... The torque distributed collaborative allocation unit connects to the series elastic actuator resonant dynamic suppression unit and uses a multi-joint torque collaborative allocation algorithm to weight the torque data. The robot cluster collaborative control command generation unit communicates with both the multi-joint torque distributed collaborative allocation unit and the humanoid robot joint torque multi-dimensional sensing and acquisition unit, and generates and corrects torque control commands based on feedback data. The joint actuator closed-loop drive unit receives the output signal from the robot cluster collaborative control command generation unit, drives the joint to complete the specified action, and simultaneously feeds back the execution status data to the humanoid robot joint torque multi-dimensional sensing and acquisition unit, forming a complete control closed loop. All units synchronize data transmission and work collaboratively through a high-speed bus to ensure the real-time performance and accuracy of torque control.

[0015] Beneficial Effects: This invention proposes a joint torque sensing chip control method and system for humanoid robots. By constructing a deep integration architecture of sensing chips, multi-joint torque adjustment mechanisms, and cluster control platforms, it completely breaks through the limitations of traditional decentralized operation of various modules. It achieves full-link dynamic linkage of data acquisition, torque deduction, interference suppression, collaborative allocation, command generation, and closed-loop correction, effectively eliminating the timing deviation between data transmission and command execution, and enabling real-time matching between torque adjustment parameters and joint dynamic states. This successfully overcomes the shortcomings of existing technologies where multiple links cannot be linked for control. Meanwhile, addressing the issue of insufficient adaptability between sensor chip-collected data and subsequent control processes, a dedicated data processing and parameter calibration logic tailored to joint dynamic changes is established. In multi-robot collaborative scenarios, a unified torque distribution coordination system is constructed, significantly reducing control response lag, improving the balance and rationality of torque distribution, fully releasing the real-time sensing capabilities of the sensor chip and the adjustment potential of the control system, significantly optimizing the accuracy, dynamic adaptability, and multi-unit collaboration of humanoid robot joint control, fully meeting the high-performance requirements of robots for joint torque control in complex work scenarios, and effectively solving the technical bottleneck that traditional control modes cannot adapt to diverse motion states. Attached Figure Description

[0016] Figure 1This is a flowchart illustrating the overall process of the method of the present invention. Figure 2 This is a flowchart of method step S3 of the present invention; Figure 3 This is a flowchart of method step S4 of the present invention; Figure 4 This is a flowchart of step S5 of the method of the present invention; Figure 5 This is a diagram showing the system unit composition of the present invention. Detailed Implementation

[0017] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] like Figure 1 As shown, the control method for a humanoid robot joint torque sensing chip includes the following steps: S1, collecting real-time torque sensing data, joint rotation displacement data, and joint motion acceleration data of each joint through the humanoid robot joint torque sensing chip, and transmitting the collected data to the robot cluster collaborative intelligent control platform for data parsing and correlation mapping; S2, constructing a flexible joint torque recursive model based on the parsed joint data, and using this model to perform hierarchical deduction of the dynamic changes in the joint torque transmission path to determine the transmission characteristics and coupling relationship of joint torque in the multi-link structure; S3, using a series elastic actuator resonance algorithm to dynamically control the resonance interference signal during the joint torque transmission process. The system employs several mechanisms: S1) Dynamic suppression, which involves adjusting the stiffness matching parameters of the actuator's elastic element through an algorithm to achieve a dynamic balance between torque transmission and resonance suppression; S4) Distributing the derived joint torque data using a multi-joint torque collaborative allocation algorithm, determining the torque allocation weight based on the motion state and load requirements of each joint; S5) Transmitting the allocated torque control signal to the humanoid robot's joint actuators through a robot cluster collaborative intelligent control platform, and dynamically correcting the torque control parameters by combining real-time feedback data from the sensor chip; and S6) Driving the joint actuators to complete the specified motion action based on the corrected control parameters, while continuously collecting joint torque response data through the sensor chip to form a closed-loop control link.

[0019] Step S1 utilizes a humanoid robot joint torque sensing chip to collect multi-dimensional data. The chip employs a 16-bit high-precision analog-to-digital converter channel with a sampling frequency set to 2000 Hz. It simultaneously collects real-time torque sensing data, joint angular displacement data, and joint acceleration data for each joint. The torque sensing data acquisition range covers a dynamic range of 0 to 500, the angular displacement data acquisition accuracy is controlled to a minimum resolution unit of 0.01, and the acceleration data acquisition response time does not exceed 1 millisecond. During data acquisition, the sensing chip establishes a data transmission link with the robot cluster's collaborative intelligent control platform via a high-speed serial communication interface. The communication baud rate is set to 115200 bits per second, and differential signal transmission is used to reduce interference. The collected raw data is encapsulated according to a preset data frame format. Each data frame includes 32 bytes of valid information, including key fields such as joint number, acquisition timestamp, torque value, angular value, and acceleration value. A cyclic redundancy check mechanism is used for error detection during data transmission, with a check bit length of 16 bits to ensure data integrity. After the data is transmitted to the control platform, the platform starts the data parsing program, splits the fields according to the data frame format, extracts the valid data and performs correlation mapping, binds the torque, rotation angle and acceleration data of the same joint at the same time stamp, establishes a three-dimensional data correlation matrix, and provides structured data support for subsequent model construction and algorithm operation. The entire data acquisition and parsing process is kept synchronized to ensure the temporal consistency and spatial correlation of the data of each joint.

[0020] Step S2 constructs a recursive model of flexible joint torque based on the structured joint data parsed in S1. During implementation, historical data sequences for each joint are first extracted, including 1000 consecutive sets of torque, rotation angle, and acceleration correlation data. The data is segmented using a sliding window method, with a window length of 50 sets and a step size of 10 sets. Trend analysis and feature extraction are performed on the data within each window to determine the dynamic law of joint torque variation with rotation angle and the influence of acceleration on torque transmission. Physical parameters such as joint link length, joint clearance, and elastic element characteristics are incorporated into the model construction. The joint link length parameter is a fixed value determined based on the robot's design dimensions. The joint clearance parameter is determined to a baseline value through multiple static calibrations. The elastic element characteristic parameters are obtained by fitting previous experimental data. The model adopts a hierarchical deduction structure, starting from the base joint and sequentially deducing torque transmission towards the end joints. The motion state parameters of the previous layer's joints are introduced as input during each deduction, while considering the coupling effect between joints. Accurate characterization of torque transmission characteristics is achieved by dynamically adjusting the deduction weights. During the simulation, the control platform updates the model input data at a frequency of 100 Hz and corrects the simulation results in real time. By comparing the deviation between the simulated torque value and the actual collected torque value, the internal parameters of the model are dynamically adjusted. The deviation threshold is set to 5. When the deviation exceeds the threshold, the parameter calibration process is initiated to ensure the accuracy of the model simulation. Finally, the model clarifies the transmission path, attenuation law and coupling strength between joint torques in the multi-link structure, providing a theoretical basis for subsequent torque control.

[0021] Step S3 employs a series elastic actuator resonance algorithm to dynamically suppress resonance interference during joint torque transmission. First, the torque data output from step S2 is frequency-decomposed using a spectrum analysis tool, with the analysis range set to 0 to 500 Hz. The characteristic frequencies of resonance interference are identified, typically concentrated in the 50 to 150 Hz range. Simultaneously, the amplitude and phase information of the resonance signal are recorded. Based on the identified resonance characteristics and combined with the structural parameters of the series elastic actuator, including the basic stiffness value of the elastic element, the actuator's natural frequency, and the initial value of the damping coefficient, the stiffness matching parameters and damping feedback parameters in the algorithm are adjusted to construct a targeted resonance suppression parameter set. This parameter set includes 10 key adjustment parameters, each with an adjustment step size of 0.01. The parameter combination is optimized using the gradient descent method. The optimized parameter set is input into the series elastic actuator resonance algorithm. The algorithm processes the torque data in real time at a computational frequency of 500 Hz, suppressing resonance interference through phase compensation and amplitude cancellation. During processing, the frequency characteristics of the torque data are monitored in real time. When the resonant characteristic frequency shifts, the parameter set update process is automatically triggered, with an update period set to 20 milliseconds. Meanwhile, the suppressed torque data is collected by the joint torque sensing chip of the humanoid robot and compared with the data before suppression to calculate the resonance suppression efficiency. When the suppression efficiency is lower than 90%, the secondary optimization process is started to adjust the algorithm operation parameters to ensure that the amplitude of resonance interference during torque transmission is reduced to less than 10% of the original amplitude, so as to achieve a dynamic balance between torque transmission and resonance suppression and ensure the stability and accuracy of torque data.

[0022] Step S4 employs a multi-joint torque collaborative allocation algorithm to distribute the joint torque data processed in Step S3. During implementation, real-time operational status data for each joint is first acquired through a robot swarm collaborative intelligent control platform. This data includes the joint's current load value, movement speed, cumulative running time, and temperature parameters. The load value is collected at 10-millisecond intervals, speed data is calculated using differential calculations of angular displacement data, and the temperature parameter range is -40 to 85 degrees Celsius. Based on the acquired status data, a joint status evaluation matrix is ​​established, with a dimension equal to the number of joints multiplied by 4. A weighted summation method is used to calculate the comprehensive status score for each joint, with weights allocated as follows: load value 30%, movement speed 25%, cumulative running time 25%, and temperature parameter 20%. The score range is set from 0 to 100 points. Based on the score results, joints are classified into three priority levels: high priority, medium priority, and low priority. The torque allocation weight coefficients for each joint are determined based on their priority levels. High-priority joints have a weight coefficient ranging from 0.6 to 0.8, medium-priority joints from 0.3 to 0.5, and low-priority joints from 0.1 to 0.2. Simultaneously, the load adaptation coefficient for each joint is considered; this coefficient is dynamically adjusted based on the ratio of the load value to the rated load, with an adjustment range of 0.8 to 1.2. The weight coefficients, load data, and torque data output from step S3 are input into a multi-joint torque collaborative allocation algorithm. The algorithm employs a distributed computing architecture, with each joint allocated an independent computing thread. The computing cycle is set to 15 milliseconds. Dynamic allocation of torque data is achieved through iterative calculation. After allocation, the target torque value for each joint is output. Simultaneously, parameter adjustment data and allocation results are recorded during the allocation process, forming an allocation log. This log provides data support for subsequent control optimization, ensuring that the torque allocation of each joint matches its own state and overall motion requirements.

[0023] Step S5 transmits the torque control signals allocated in Step S4 to the humanoid robot's joint actuators via a robot cluster collaborative intelligent control platform. During implementation, the allocated target torque value is first encoded using an 8-bit binary encoding method, with each code corresponding to a unique torque control level. A check bit is added during encoding to ensure signal transmission accuracy. The control signals are transmitted to the drive modules of each joint actuator via an industrial Ethernet interface, with transmission latency controlled within 5 milliseconds. Upon receiving the signal, the drive module decodes it to reconstruct the target torque value and simultaneously activates the high-frequency data acquisition mode of the humanoid robot's joint torque sensor chip, increasing the acquisition frequency to 3000 Hz to collect the actual torque response data of the joint actuators in real time. The actual torque response data is compared with the target torque value, and the deviation is calculated using the absolute difference method. When the deviation is less than 3, the current control parameters remain unchanged; when the deviation is between 3 and 10, a small correction mode is activated, with a correction magnitude of 10% of the deviation; when the deviation is greater than 10, a large correction mode is activated, with a correction magnitude of 20% of the deviation. During the correction process, real-time motion data of the joint is referenced, including the rate of change of rotation angle and the trend of acceleration change, and the torque control parameters are dynamically adjusted. The adjusted parameters are transmitted to the drive module through the feedback link. The drive module updates the control signal output according to the new parameters. The entire correction process forms a closed-loop control. The correction period is set to 10 milliseconds to ensure that the actual torque response data continuously approaches the target torque value, thereby improving the accuracy and dynamic response capability of torque control.

[0024] Step S6, based on the control parameters corrected in step S5, drives the joint actuator to complete the specified motion. During implementation, the drive module outputs a corresponding drive current signal according to the corrected torque control parameters. The adjustment accuracy of the current signal is controlled within 0.01 amperes. The signal is amplified to the power level required by the actuator through a power amplifier circuit, driving the motor to move the joint linkage. During the motion, the humanoid robot's joint torque sensor chip continuously collects joint torque response data at a frequency of 2000 Hz, while simultaneously recording the actual angular position and acceleration data of the joint. This data is fed back to the robot cluster collaborative intelligent control platform in real time via a high-speed data transmission link. The transmission process uses data packet grouping, with each data packet containing 100 sets of sampled data to ensure efficient and complete data transmission. The control platform monitors and analyzes the feedback data in real time to determine whether the joint motion state is consistent with the specified action requirements. The analysis indicators include torque response stability, angular position accuracy, and acceleration change smoothness. The allowable error range for angular position accuracy is ±0.02, and the acceleration change smoothness is evaluated by calculating the acceleration difference between adjacent sampling points, with a difference threshold set at 0.5. When the monitored data exceeds the allowable range, the platform immediately triggers a control parameter fine-tuning command. The fine-tuning range is 5% of the current parameter. This is quickly transmitted to the drive module through the closed-loop control link to correct the drive signal in real time, ensuring that the joint actuator always completes the motion action according to the specified trajectory. At the same time, it continuously collects and stores all data in the entire motion process to form a complete motion control data archive, providing comprehensive data support for subsequent control strategy optimization and system performance improvement.

[0025] Preferably, the recursive model expression for the flexible joint torque is: ,in, For the first The recursive torque value of each joint, Let be the initial torque value of the i-th joint. Let be the rotation angle of the i-th joint. Let be the stiffness coefficient of the series elastic actuator. For the deformation of the elastic element, Let be the rotational inertia of the i-th joint. Let be the angular acceleration of the i-th joint. This is the torque transmission coupling coefficient.

[0026] Specifically, the flexible joint torque recursive model is based on the mechanical transmission characteristics of flexible joints and the principle of multi-joint coupling. First, torque transmission data under different joint angles and elastic element deformations are measured experimentally, accumulating 500 sets of valid samples. The linear correlation between torque transmission and the cosine of the angle, the sine of the deformation, and angular acceleration is analyzed. Simultaneously, the influence of joint rotational inertia on transmission efficiency is considered, and a coupling coefficient is introduced to balance multi-joint linkage errors. When establishing the model, the torque transmission of adjacent joints is used as the core link, with the initial torque of the previous joint as the basic input. The torque attenuation effect of the angle, the additional torque generated by the elastic element deformation, and the inertial torque caused by angular acceleration are combined, and a complete recursive relationship is constructed through linear superposition. The rotational inertia is obtained by calculating the volume and density distribution of the joint entity using 3D modeling software. The coupling coefficient is determined after 100 iterations of calibration, with a value ranging from 0.85 to 0.98. The stiffness coefficient of the elastic element is obtained through tensile testing experiments and fixed as the measured average value. During implementation, the control platform calls the rotation angle and acceleration data and preset physical parameters collected in step S1 in real time, substitutes them into the model to perform hierarchical torque extrapolation, updates the coupling coefficient once for each joint extrapolation, and keeps the extrapolation frequency consistent with the data acquisition frequency. This model accurately describes the transmission path and attenuation law of torque in multi-link structures, providing a precise theoretical basis for subsequent torque control and solving the extrapolation error problem caused by neglecting the joint coupling effect in traditional models.

[0027] Preferably, the expression for the series elastic actuator resonance algorithm is: ,in, The transfer function after resonance suppression. The resonant frequency, For complex frequency variables, The damping coefficient is... For damping feedback coefficient, For the moment of inertia of the motor, This is the motor damping coefficient. This is the frequency compensation factor.

[0028] Specifically, the resonance algorithm for the series elastic actuator is based on the dynamic equations of the series elastic actuator and combines the phase compensation principle for resonance suppression. It uses a spectrum analysis tool to decompose the vibration data during actuator operation, identifying the amplitude and phase characteristics within the resonant frequency range, and establishing the numerator and denominator structure of the transfer function. The numerator considers the suppression effect of elastic stiffness and damping coefficient on resonance, while the denominator incorporates the rotational inertia, damping coefficient, and stiffness parameters of the motor and load, constructing a fourth-order transfer function to cover resonance interference over a wide frequency range. The motor's rotational inertia and damping coefficient are determined through experimental calibration using parameters from the motor nameplate. The elastic stiffness and damping feedback coefficient are optimized through 20 sets of resonance suppression experiments at different frequencies. The frequency compensation factor is set according to the resonant frequency offset range, with a value between 0.9 and 1.1. In implementation, the resonant characteristic frequency is first extracted from the torque data output in step S2. Based on this frequency, the parameter combination in the transfer function is adjusted, and the gradient descent method is used for iterative optimization 100 times to determine the optimal parameter set. The algorithm operation frequency is set to 500 Hz, and the torque data is filtered in real time. Phase compensation is used to offset the phase deviation of the resonant signal, and amplitude suppression is used to reduce the interference intensity, ensuring that the amplitude of resonant interference is controlled within the allowable range during torque transmission. The establishment of this model solves the torque fluctuation problem caused by the inherent resonance of the series elastic actuator and improves the stability of torque transmission.

[0029] Preferably, the expression for the multi-joint torque collaborative allocation algorithm is: ,in, The distributed joint torque, The total number of joints. Assign weighting coefficients to the torque of the i-th joint. This is the joint motion sensitivity coefficient. Let be the generalized coordinate of the i-th joint. For generalized coordinate changes, For load adaptability coefficient, Let be the load torque of the i-th joint. To coordinate the allocation of coordination factors.

[0030] Specifically, the multi-joint torque collaborative allocation algorithm is based on the principle of multi-objective optimization, comprehensively considering the load-bearing capacity, motion sensitivity, and load adaptation requirements of each joint. It determines the weight ratio of each influencing factor through the analytic hierarchy process (AHP) and constructs the torque allocation relationship by combining a distributed allocation strategy. In the derivation process, a joint state evaluation index system is first established, including dimensions such as load, motion speed, and health status. Through normalization, each index is transformed into a unified scoring standard. Then, torque weight coefficients are allocated based on the scoring results. Simultaneously, a generalized coordinate change is introduced to characterize the joint motion trend, and the load torque reflects the actual force situation. Dynamic torque allocation is achieved through linear combination. The weight coefficients are set according to the joint priority level, and the load adaptation coefficient is dynamically adjusted based on the ratio of the load value to the rated load, ranging from 0.8 to 1.2. The collaborative allocation coordination factor, after multiple cluster control experiments, is fixed at 0.95. During implementation, the control platform collects load, speed, and temperature data of each joint in real time, calculates the comprehensive status score and determines the corresponding parameters, and substitutes them into the algorithm for distributed computation. Each joint is assigned an independent computation thread, and the computation cycle is set to 15 milliseconds. The target torque value of each joint is obtained through iterative calculation, and the parameter adjustment trajectory is recorded at the same time. This model realizes the balanced distribution of torque of multiple joints, avoids the situation of overload or insufficient load of a single joint, and improves the coordination and stability of robot movement.

[0031] Preferably, the torque control parameter correction model expression of the robot cluster collaborative intelligent control platform is as follows: To control the parameter correction amount, This is the proportionality coefficient. The integral coefficient is... The differential coefficients are... For coupling correction coefficients, For reference torque value, This is the feedback torque value of the sensor chip. This is a multi-joint coupling torque.

[0032] Specifically, the torque control parameter correction model of the robot swarm collaborative intelligent control platform combines the proportional-integral-derivative (PI-DE) control principle with a multi-joint coupling torque compensation mechanism. By analyzing the impact of control deviation, deviation integral, and deviation derivative on the control effect, the weighted combination relationship of these three factors is determined. Simultaneously, considering the interference of coupling torque generated by multi-joint linkage on control accuracy, a coupling correction coefficient is introduced for error compensation. During the derivation process, 1000 sets of data corresponding to control deviation and parameter correction amounts were collected. The optimal range of proportional, integral, and derivative coefficients was obtained through linear regression analysis. The coupling correction coefficient was then optimized using measured data of coupling torque. The proportional coefficient ranges from 0.3 to 0.6, the integral coefficient from 0.05 to 0.15, and the derivative coefficient from 0.1 to 0.2. The coupling correction coefficient is dynamically adjusted according to the number of linked joints; the more linked joints, the larger the coefficient, with a maximum value not exceeding 0.3. During implementation, the control platform calculates the deviation between the reference torque and the feedback torque in real time, along with the cumulative integral deviation and the rate of change of the derivative deviation. Combined with the detected coupling torque data, these are substituted into the model to calculate the control parameter correction. The correction period is consistent with the closed-loop control period. The proportional element responds quickly to the deviation, the integral element eliminates static errors, the derivative element suppresses overshoot, and the coupling correction coefficient compensates for multi-joint linkage interference. This ensures that the control parameters adapt to the joint motion state in real time, significantly improving the accuracy and dynamic response capability of torque control.

[0033] Preferably, the signal acquisition model expression of the humanoid robot joint torque sensing chip is as follows: ense ,in, The chip outputs a voltage signal. ense is the sensitivity coefficient of the sensor chip. This represents the actual torque of the joint. This is the torque-to-voltage conversion factor. The rate of change of torque, For dynamic response coefficients, Zero-point offset voltage, This is the environmental compensation coefficient.

[0034] Specifically, the signal acquisition model of the humanoid robot joint torque sensing chip is based on the torque-to-voltage conversion principle of the sensing chip, combined with dynamic response characteristics and environmental interference compensation mechanisms. Through experimental measurement of the chip's output voltage under different torques and torque change rates, the linear correlation between the voltage signal and the torque and torque change rate is analyzed. Simultaneously, the influence of zero-point offset and environmental factors on the output signal is considered, and an environmental compensation coefficient is introduced to correct the error. In the derivation process, a high-precision torque loading device is used to calibrate the chip, accumulating 200 different torque values ​​and recording the corresponding output voltages. The torque-to-voltage conversion coefficient is obtained by least squares fitting. Simultaneously, 100 sets of experiments are conducted under different temperature and humidity environments to determine the value pattern of the environmental compensation coefficient. The sensor chip sensitivity coefficient is determined by combining the chip datasheet with actual measurement calibration. The torque-to-voltage conversion coefficient is a fixed value after fitting. The dynamic response coefficient is set based on the chip response time test results. The zero-point offset voltage is obtained through static calibration. The environmental compensation coefficient is adjusted based on real-time monitored environmental data, with a value range between 0.92 and 1.08. During implementation, the sensor chip collects joint torque and torque change rate data, combines the built-in zero-point offset voltage and real-time environmental parameters, and substitutes them into the model to calculate the output voltage signal. The signal acquisition frequency and the model calculation frequency are kept synchronized. Through this model, the physical torque signal is accurately converted into an electrical signal, while compensating for errors caused by environmental interference and zero-point offset, ensuring the accuracy and stability of the output signal, and providing a reliable signal source for subsequent data processing and control command generation.

[0035] Preferred, such as Figure 2 As shown, step S3 includes the following sub-steps: S31, extracting the resonant characteristic frequency from the torque data output by the flexible joint torque recursive model, and determining the amplitude and phase characteristics of the resonant signal through spectrum analysis; S32, based on the structural parameters and motion state of the series elastic actuator, adjusting the stiffness matching parameters and damping feedback coefficients in the algorithm to construct a resonant suppression parameter set; S33, inputting the resonant characteristic parameters and suppression parameter set into the series elastic actuator resonant algorithm, and generating a resonant suppression control signal through algorithm calculation; S34, superimposing the resonant suppression control signal onto the joint torque control link to cancel the resonant interference in the torque transmission process in real time, and simultaneously collecting the suppressed torque data through a sensor chip to verify the effect.

[0036] Specifically, in step S31, the control platform calls the torque data output from step S2, starts the spectrum analysis program, sets the analysis frequency range to 0 to 500 Hz, uses the Fast Fourier Transform algorithm to decompose the data into frequencies with a decomposition accuracy controlled within 1 Hz, identifies the resonant characteristic frequency through a peak detection algorithm, and records the amplitude and phase information of the resonant signal at that frequency. The entire analysis process takes no more than 20 milliseconds, ensuring the real-time extraction of resonant features. Step S32, based on the structural parameters of the series elastic actuator, including the basic value of the elastic element stiffness, the actuator's natural frequency, and the initial value of the damping coefficient, combines the resonant features identified in S31, adjusts the stiffness matching parameters and damping feedback coefficient through a parameter optimization algorithm, and constructs a resonant suppression parameter set including 10 key adjustment parameters. The adjustment step size for each parameter is set to 0.01, and the initial parameter combination is determined through multiple iterations. Step S33 inputs the resonant characteristic parameters and suppression parameter set into the series elastic actuator resonant algorithm. The algorithm performs real-time calculations at a frequency of 500 Hz, adjusting the torque signal phase through a phase compensation algorithm and reducing the intensity of resonant interference through an amplitude suppression algorithm, generating a targeted resonant suppression control signal. The effectiveness of the parameters is monitored in real time during the calculation. Step S34 superimposes the resonant suppression control signal onto the joint torque control link via a data bus, using a linear superposition strategy. Simultaneously, the high-frequency acquisition mode of the humanoid robot's joint torque sensor chip is activated, increasing the acquisition frequency to 3000 Hz. The suppressed torque data is acquired and compared with the data before suppression. The deviation threshold is set to 5%. When the deviation exceeds the threshold, the parameters are re-optimized to ensure effective suppression of resonant interference and guarantee the stability of torque transmission.

[0037] Preferred, such as Figure 3 As shown, step S4 includes the following sub-steps: S41, acquiring real-time load data, motion speed data, and joint health status data of each joint through the robot cluster collaborative intelligent control platform, and establishing a joint status evaluation matrix; S42, determining the torque bearing capacity and motion priority of each joint based on the joint status evaluation matrix, and assigning corresponding torque allocation weight coefficients and load adaptation coefficients; S43, inputting the weight coefficients, load data, and recursive torque values ​​into the multi-joint torque collaborative allocation algorithm, and dynamically allocating torque data through the distributed operation of the algorithm; S44, outputting the allocated torque control commands for each joint, and simultaneously recording the parameter adjustment data during the allocation process to provide a basis for subsequent control optimization.

[0038] Specifically, in step S41, the multi-channel data acquisition interface of the robot cluster collaborative intelligent control platform is used to acquire the running status data of each joint in real time. The load data acquisition interval is set to 10 milliseconds. The motion speed data is calculated by differential calculation of the rotation displacement data, with a sampling period of 5 milliseconds. The joint health status data is acquired by temperature sensors and vibration sensors at a frequency of 100 Hz. These data are classified and stored according to the joint number, and a joint status evaluation matrix with a dimension of 4 times the number of joints is constructed. The matrix data update frequency is consistent with the load data acquisition interval. Step S42 uses the analytic hierarchy process (AHP) to determine the weighting of each evaluation indicator: load value (30%), movement speed (25%), cumulative running time (25%), and temperature parameter (20%). A weighted summation formula is used to calculate the comprehensive status score for each joint, ranging from 0 to 100 points. Based on the scores, joints are divided into high, medium, and low priorities. High-priority joints score no less than 80 points, medium-priority joints score 50 to 79 points, and low-priority joints score below 50 points. Corresponding torque allocation weighting coefficients of 0.6 to 0.8, 0.3 to 0.5, and 0.1 to 0.2 are assigned. The load adaptation coefficient is dynamically adjusted based on the ratio of the load value to the rated load, ranging from 0.8 to 1.2. Step S43 inputs the determined weighting coefficients, load data, and torque data output from step S3 into a multi-joint torque collaborative allocation algorithm. The algorithm uses a distributed computing architecture, allocating an independent computing thread to each joint. The computing cycle is set to 15 milliseconds, and dynamic allocation of torque data is achieved through iterative calculation, with no more than 50 iterations to ensure computational efficiency. Step S44 outputs the torque control commands for each joint after allocation. The command format is encapsulated in a preset 32-byte data frame, including information such as joint number, target torque value, and allocation timestamp. At the same time, the parameter adjustment data and allocation results during the allocation process are recorded through the data storage module to form an allocation log. The log storage period is 30 days, providing complete data support for subsequent control optimization and realizing balanced allocation of torque across multiple joints.

[0039] Preferred, such as Figure 4 As shown, step S5 includes the following sub-steps: S51, receiving the torque control command output by the multi-joint torque collaborative allocation algorithm, and determining the command execution priority by combining the global control strategy of the robot cluster collaborative intelligent control platform; S52, transmitting the torque control command to the drive module of each joint actuator through the communication link, and simultaneously starting the high-frequency data acquisition mode of the sensor chip; S53, comparing the real-time torque data fed back by the sensor chip with the command torque data, and calculating the torque control deviation value; S54, based on the deviation value and the joint motion state, dynamically correcting the torque control parameters through the control platform, and retransmitting the corrected command to the drive module for closed-loop control.

[0040] Specifically, step S51 receives torque control commands output by the multi-joint torque collaborative allocation algorithm. The command parsing module extracts key information such as the target torque value, joint number, and execution priority. Execution priorities are set according to the robot's overall motion task and are divided into three categories: urgent, normal, and low priority. Urgent priority commands directly enter the execution queue, while normal and low priority commands are queued according to the order of receipt. Simultaneously, combined with the global control strategy of the robot cluster collaborative intelligent control platform, the command execution order is dynamically adjusted with an adjustment cycle of 5 milliseconds to ensure the priority execution of critical actions. Step S52 transmits the torque control commands to the drive modules of each joint actuator via a high-speed communication bus. The communication bus uses industrial Ethernet with a transmission rate set to 100 megabits per second and a transmission delay controlled within 5 milliseconds. After receiving the commands, the drive modules decode and verify them. Upon successful verification, the high-frequency data acquisition mode of the humanoid robot's joint torque sensing chip is activated, with the acquisition frequency set to 3000 Hz. Actual joint torque data, rotational angle data, and acceleration data are collected. The collected data is stored in real-time in the cache module, with a cache capacity of 1000 sets of data. Step S53 uses the data comparison module to retrieve real-time torque data from the cache and target torque data from the command. The absolute difference method is used to calculate the deviation between the two, with the calculation cycle consistent with the acquisition cycle. The deviation trend is recorded, and a sliding window method is used to smooth the deviation data. The window length is set to 50 data sets to eliminate deviation fluctuations caused by instantaneous interference, ensuring the accuracy of the deviation value. Step S54, based on the calculated deviation value and joint motion state data, dynamically adjusts the torque control parameters through the parameter correction module. When the deviation value is less than 3, the current parameters remain unchanged; when the deviation value is between 3 and 10, a small correction of 10% is made; when the deviation value is greater than 10, a large correction of 20% is made. The corrected parameters are transmitted to the drive module in real time via the feedback link. The drive module updates the control signal output according to the new parameters, forming a closed-loop control. The correction cycle is set to 10 milliseconds to ensure that the actual torque response continuously approaches the target torque value, improving control accuracy.

[0041] like Figure 5As shown, a humanoid robot joint torque sensing chip control system is characterized by being applied to a humanoid robot joint torque sensing chip control method, comprising: a humanoid robot joint torque multi-dimensional sensing and acquisition unit, used to collect torque data, angular displacement data, and acceleration data of each joint, and establish a bidirectional data transmission link with a robot cluster collaborative intelligent control platform; a flexible joint torque hierarchical recursive calculation unit, connected to the humanoid robot joint torque multi-dimensional sensing and acquisition unit, receiving the collected data and performing torque transmission characteristic deduction through a preset flexible joint torque recursive model; and a series elastic actuator resonance dynamic suppression unit, connected to the flexible joint torque hierarchical recursive calculation unit, suppressing resonance interference in the torque data through a series elastic actuator resonance algorithm. A multi-joint torque distributed collaborative allocation unit, connected to a series elastic actuator resonant dynamic suppression unit, utilizes a multi-joint torque collaborative allocation algorithm to weightedly allocate torque data. A robot cluster collaborative control command generation unit communicates with both the multi-joint torque distributed collaborative allocation unit and the humanoid robot joint torque multi-dimensional sensing and acquisition unit, generating and correcting torque control commands based on feedback data. A joint actuator closed-loop drive unit receives the output signal from the robot cluster collaborative control command generation unit, drives the joints to complete specified actions, and simultaneously feeds back execution status data to the humanoid robot joint torque multi-dimensional sensing and acquisition unit, forming a complete control closed loop. All units synchronize data transmission and collaborate via a high-speed bus to ensure the real-time performance and accuracy of torque control. The humanoid robot joint torque sensing chip control method and system construct a deeply integrated architecture of the sensing chip, multi-joint torque adjustment mechanism, and cluster control platform, breaking down the barriers of decentralized and independent operation of control modules in traditional technologies. By integrating the entire process of data acquisition, torque calculation, interference suppression, coordinated allocation, command generation, and closed-loop correction, dynamic linkage and synchronous data transmission between each stage are achieved. This effectively eliminates the timing deviation between data transmission and command execution in traditional modes, enabling torque adjustment parameters to adapt to the dynamic state of the joint in real time. This design precisely overcomes the core deficiency of existing technologies where multiple stages cannot be linked for control, transforming joint torque control from decentralized adjustment to integrated collaborative control. This significantly improves the coherence and synchronization of the control process, laying the foundation for the precision of joint control.

[0042] The formulas in this invention incorporate different scalar and vector parameters into the same calculation system. Adaptation logic is established through physical meaning association and dimensional unification. First, the essential role and mutual influence of each parameter in joint torque control are clarified. Then, dimensional differences are eliminated through parameter standardization. Taking the recursive formula for flexible joint torque as an example, scalar parameters such as the stiffness of elastic elements and moment of inertia reflect the inherent properties of the joint structure, while vector parameters such as joint angle and acceleration characterize the direction and magnitude of the motion state. When constructing the formula, the mechanism of action of scalars on vectors is first established through mechanical principles. Then, the directional information of vector parameters is converted into scalar coefficients (such as the cosine and sine values ​​of the angle), thus forming a quantitative correlation between the directionality of the vector and the numerical value of the scalar. In parameter processing, the adaptation coefficients between scalars and vectors are determined through experimental calibration. The magnitudes of vector parameters and scalar parameters are scaled to the same order of magnitude. For example, the magnitude of the acceleration vector and the scalar of moment of inertia are adjusted to the same numerical range. Finally, they are integrated through linear superposition, multiplication, and other operations. Meanwhile, the formula introduces adjustment parameters such as coupling coefficients and compensation factors to balance the weights of scalar and vector actions, ensuring that the calculation results of different types of parameters conform to the physical laws of torque control. This design is based on the inherent correlation between various parameters in torque transmission, resonance suppression, and collaborative allocation. Through dimensional unification, direction quantization, and weight adaptation, it achieves the fusion calculation of heterogeneous parameters, preserving the essential characteristics of each parameter while forming a logically self-consistent calculation system, thus meeting the needs of humanoid robot joint torque control for multi-dimensional parameter collaborative processing.

[0043] This invention optimizes the design of sensor data processing and collaborative allocation, establishing dedicated data processing and parameter calibration logic that adapts to the dynamic changes of joints, and constructing a unified torque allocation coordination system for multi-robot collaborative scenarios. The dedicated data processing logic enhances the adaptability between sensor chip-acquired data and subsequent control processes, avoiding information loss and adaptation deviations during data transmission. The unified torque allocation coordination system standardizes the torque allocation rules among multiple joints and robots, significantly reducing control response lag and improving the balance and rationality of torque allocation. This design successfully solves the shortcomings of insufficient adaptability between sensor data and control processes and disordered torque allocation in collaborative scenarios in existing technologies. It fully releases the real-time sensing capabilities of sensor chips and the adjustment potential of the control system, significantly optimizing the dynamic adaptability of joint control and multi-unit collaboration, and meeting the high-performance control requirements in complex operating scenarios.

[0044] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0045] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for controlling the joint torque sensing chip of a humanoid robot, characterized in that, Includes the following steps: S1 collects real-time torque sensing data, joint rotation displacement data, and joint motion acceleration data of each joint through the joint torque sensing chip of the humanoid robot, and transmits the collected data to the robot cluster collaborative intelligent control platform for data parsing and correlation mapping; S2. Based on the analyzed joint data, a flexible joint torque recursive model is constructed. Through this model, the dynamic changes in the joint torque transmission path are hierarchically deduced to determine the transmission characteristics and coupling relationship of joint torque in multi-link structures. S3 employs a series elastic actuator resonance algorithm to dynamically suppress resonance interference signals during joint torque transmission. The algorithm adjusts the stiffness matching parameters of the actuator's elastic element to achieve a dynamic balance between torque transmission and resonance suppression. S4. The multi-joint torque collaborative allocation algorithm is used to distribute the simulated joint torque data in a distributed manner, and the torque allocation weight is determined according to the motion state and load requirements of each joint. S5 transmits the allocated torque control signals to the humanoid robot joint actuators through the robot cluster collaborative intelligent control platform, and dynamically corrects the torque control parameters by combining real-time feedback data from the sensor chip. S6 drives the joint actuator to complete the specified motion action based on the corrected control parameters, and at the same time continuously collects joint torque response data through the sensor chip to form a closed-loop control link.

2. The humanoid robot joint torque sensing chip control method according to claim 1, characterized in that, The recursive model expression for the flexible joint torque is as follows: ， in, For the first The recursive torque value of each joint, Let be the initial torque value of the i-th joint. Let be the rotation angle of the i-th joint. Let be the stiffness coefficient of the series elastic actuator. For the deformation of the elastic element, Let be the rotational inertia of the i-th joint. Let be the angular acceleration of the i-th joint. This is the torque transmission coupling coefficient.

3. The humanoid robot joint torque sensing chip control method according to claim 1, characterized in that, The expression for the series elastic driver resonance algorithm is as follows: ， in, The transfer function after resonance suppression. The resonant frequency, For complex frequency variables, The damping coefficient is... For damping feedback coefficient, For the moment of inertia of the motor, This is the motor damping coefficient. This is the frequency compensation factor.

4. The humanoid robot joint torque sensing chip control method according to claim 1, characterized in that, The expression for the multi-joint torque collaborative allocation algorithm is as follows: ， in, The distributed joint torque, The total number of joints. Assign weighting coefficients to the torque of the i-th joint. This is the joint motion sensitivity coefficient. Let be the generalized coordinate of the i-th joint. For generalized coordinate changes, For load adaptability coefficient, Let be the load torque of the i-th joint. To coordinate the allocation of coordination factors.

5. The humanoid robot joint torque sensing chip control method according to claim 1, characterized in that, The torque control parameter correction model expression of the robot cluster collaborative intelligent control platform is as follows: ; To control the parameter correction amount, This is the proportionality coefficient. The integral coefficient is... The differential coefficients are... For coupling correction coefficients, For reference torque value, This is the feedback torque value of the sensor chip. This is a multi-joint coupling torque.

6. The humanoid robot joint torque sensing chip control method according to claim 1, characterized in that, The signal acquisition model expression for the humanoid robot joint torque sensing chip is as follows: nape , in, The chip outputs a voltage signal. ense is the sensitivity coefficient of the sensor chip. This represents the actual torque of the joint. This is the torque-to-voltage conversion factor. The rate of change of torque, For dynamic response coefficients, Zero-point offset voltage, This is the environmental compensation coefficient.

7. The humanoid robot joint torque sensing chip control method according to claim 1, characterized in that, S3 includes the following steps: S31, extract the resonant characteristic frequency from the torque data output by the flexible joint torque recursive model, and determine the amplitude and phase characteristics of the resonant signal through spectrum analysis; S32, based on the structural parameters and motion state of the series elastic actuator, adjust the stiffness matching parameters and damping feedback coefficients in the algorithm to construct a set of resonance suppression parameters; S33, input the resonant characteristic parameters and the suppression parameter set into the series elastic actuator resonant algorithm, and generate the resonant suppression control signal through algorithm calculation; S34 superimposes the resonance suppression control signal onto the joint torque control link to cancel the resonance interference in the torque transmission process in real time, and at the same time, the suppressed torque data is collected by the sensor chip to verify the effect.

8. The humanoid robot joint torque sensing chip control method according to claim 1, characterized in that, S4 includes the following steps: S41, through the robot cluster collaborative intelligent control platform, obtains real-time load data, motion speed data and joint health status data of each joint, and establishes a joint status assessment matrix; S42, Based on the joint state evaluation matrix, determine the torque bearing capacity and motion priority of each joint, and assign the corresponding torque allocation weight coefficient and load adaptation coefficient. S43, input the weighting coefficient, load data and recursive torque value into the multi-joint torque collaborative allocation algorithm, and dynamically allocate the torque data through the distributed operation of the algorithm; S44 outputs the torque control commands for each joint after allocation, and records the parameter adjustment data during the allocation process to provide a basis for subsequent control optimization.

9. The control method for the joint torque sensing chip of a humanoid robot according to claim 1, characterized in that, S5 includes the following steps: S51 receives torque control commands output by the multi-joint torque collaborative allocation algorithm, and determines the command execution priority by combining the global control strategy of the robot cluster collaborative intelligent control platform. S52 transmits torque control commands to the drive modules of each joint actuator via a communication link, and simultaneously activates the high-frequency data acquisition mode of the sensor chip. S53 compares the real-time torque data fed back by the sensor chip with the command torque data to calculate the torque control deviation value; S54, based on the deviation value and joint motion state, dynamically corrects the torque control parameters through the control platform, and retransmits the corrected command to the drive module for closed-loop control.

10. A joint torque sensing chip control system for a humanoid robot, characterized in that, This system is applied to the humanoid robot joint torque sensing chip control method according to claim 1, comprising: The humanoid robot joint torque multi-dimensional sensing and acquisition unit is used to collect joint torque data, rotational displacement data and acceleration data, and establish a two-way data transmission link with the robot cluster collaborative intelligent control platform; The flexible joint torque hierarchical recursive calculation unit is connected to the humanoid robot joint torque multi-dimensional sensing and acquisition unit. It receives the acquired data and performs torque transmission characteristic deduction through the preset flexible joint torque recursive model. The series elastic actuator resonance dynamic suppression unit and the flexible joint torque hierarchical recursive calculation unit suppress resonance interference in torque data through the series elastic actuator resonance algorithm. A multi-joint torque distributed collaborative allocation unit is connected to a series elastic actuator resonant dynamic suppression unit, and a multi-joint torque collaborative allocation algorithm is used to perform weighted allocation of torque data. The robot cluster collaborative control command generation unit communicates with the multi-joint torque distributed collaborative allocation unit and the humanoid robot joint torque multi-dimensional sensing and acquisition unit, respectively, and generates and corrects torque control commands by combining feedback data. The joint actuator closed-loop drive unit receives the output signal from the robot cluster collaborative control command generation unit, drives the joint to complete the specified action, and feeds back the execution status data to the humanoid robot joint torque multi-dimensional sensing and acquisition unit to form a complete control closed loop. Each unit transmits data synchronously and works collaboratively through a high-speed bus to ensure the real-time performance and accuracy of torque control.

Images